Wavelength agility enables optical networking

Robert Plastow

Wavelength-tunable lasers are moving rapidly to become a mainstream component in optical fiber networks. These new components allow carriers to provision bandwidth flexibly and affordably. Not only do they provide immediate cost savings in applications such as long-haul sparing and inventory control, but they also allow new system architectures in which wavelength channels can be provided when and where needed. This is a key requirement because carriers will increasingly make money from the ability to add and drop traffic as the customer demands (see "Tunable lasers route optical signals," p. 26).

Tunable lasers will first be used in sparing for long-haul networks. This simple application replaces spare transmitter cards of up to 80 different wavelengths with a single tunable spare. Savings in inventory also can be achieved using these same transmitters, and this will be an increasing driver as volumes increase.

In addition, packet transmission of data by wavelength makes the most efficient use of optical networks. Recent work in experimental test beds has demonstrated that tunable lasers can switch wavelength on a packet-by-packet basis.1

Approaches to tunability

There are three basic approaches to making a laser tunable. They all involve changing the optical path length within the lasing cavity.

Mechanical tuning. Tunable lasers used in instrumentation have traditionally been based on a movable external wavelength-selective mirror, normally a diffraction grating, and a semiconductor gain section. These structures have the advantage that, with sufficiently complex mechanical movements, the tuning can be truly continuous over a large wavelength range.

The long cavity also gives a narrow linewidth-in the kHz region. However, the bulk, cost, complexity, and-most important-limited reliability of this approach make it unsuitable for telecommunications. More recently, microelectromechanical (MEMS) structures have been demonstrated, including vertical-cavity surface-emitting lasers (VCSELs) with moving mirrors. Although promising, this technology is not yet available in a volume commercial product for the 1550-nm region.

Temperature tuning. To ensure telecommunications reliability, it is desirable that the tuning mechanism be incorporated in the semiconductor device itself and that the laser use standard processes where possible. An extreme version of this approach is to take a standard distributed-feedback (DFB) laser and simply vary the refractive index of the laser cavity by changing the temperature of the laser chip. This method of variation has the attraction of working with an existing component. The disadvantages are that the tuning range is very small and that the tuning mechanism is inherently slow.

The tuning coefficient with temperature is approximately 0.08 nm/°C, requiring temperature excursions of 25°C simply to tune 2 nm. This tuning mechanism is also inherently slow, depending on temperature stabilization within the package, which makes it unsuitable for fast-packet-switching applications. It is possible to increase the tuning range by placing multiple lasers in series on a single chip and operating some at transparency and some at lasing. An approach of this type moves away from a standard device without greatly increasing the tuning range and does not address the switching speed.

Current tuning. A more efficient way to alter the refractive index is by changing the current, and hence the carrier density, in the semiconductor. A DFB structure makes this impossible because the lasing and wavelength selection sections are the same. Indeed, the residual effect is the undesirable chirp, or wavelength change during modulation, that DFB designers try to avoid. However, by separating out the light generation and wavelength selection regions, it is possible to have broad wavelength tuning and low-chirp operation.

Designs using a distributed Bragg reflector (DBR) in a passive waveguide section can obtain results. These designs can have two, three, or four sections on a single chip. If only two sections are used (gain and Bragg grating) it is necessary to match the cavity mode to the Bragg wavelength by small temperature tuning of the whole cavity. A third current-tuned, passive section-the so-called phase section-removes this requirement.

By incorporating a fourth section that can provide coarse tuning over the entire C-band and modifying the grating section to provide multiple wavelength reflections over the entire C-band, it is possible to make a single laser chip-a grating-coupled sampled reflector (GCSR)-that can tune over the entire C-band. All of these designs retain the standard processes and standard packaging techniques familiar to the DFB laser, allowing telecommunications reliability to be assessed and demonstrated.

Evaluating performance

In assessing which type of tunable laser to use, the first area to address is tuning range. Modest savings in inventory can be achieved by temperature-tuning existing DFB lasers. Temperature tuning is applicable for up to four 50-GHz-spaced channels, or a total tuning range of approximately 1.6 nm. Although it is the simplest approach, it is of little use for wavelength routing and does not address the sparing or inventory issues associated with 80 or more separate channels. The temperature-tuning range is also often used to achieve the original wavelength specification, so the range of tuning available can be less.

Greater tuning range is of benefit in most applications, and here the major consideration is the desired tuning range and optical power. Three-section DBR lasers can readily achieve tuning in the 6- to 10-nm range, and the four-section GCSR design offers full C-band (32 nm) tuning from a single device. Both designs can move to greater than +10 dBm (10 mW), but the shorter three-section device will always be ahead on power by approximately 3 dB because of reduced chip losses. For sparing and add/drop in long-haul applications, the higher power is generally preferred, whereas for wavelength routing in long-haul and for most metro applications, full C-band tuning is desired.

Although the DBR and GCSR designs are capable of direct modulation with low chirp at OC-48 rates, existing products are aimed at external modulation at OC-192 rates and above. Important parameters include relative intensity noise, sidemode suppression, and polarization extinction ratio. In all of these, both designs are very similar in performance to the DFB lasers they replace.

Wavelength control

The complexity of controlling a laser with multiple parameters has required tunable lasers to be provided in a compact module with control electronics. This arrangement allows the laser to be pre-characterized and requires only digital input instructions to select a specific wavelength. It also allows different types of laser chip such as narrow-band tunable, higher power three-section DBR, and full C-band GCSR lasers to be fully interchangeable as desired (see Fig. 1).

The combination of a tunable laser and an external modulator provides a transmitter capable of any wavelength up to full C-band coverage and any data rate up to OC-192 or beyond. By separating the components, rather than creating a hybrid integration, this flexible combination is also compatible with fixed-wavelength lasers using external modulation (see Fig. 2).

To achieve this coverage, it is necessary to connect the laser and the modulator with polarization-preserving fiber, with the polarization aligned to the slow axis of the fiber. This alignment can be maintained in test by using keyed FC-PC connectors, with the key also aligned to the slow axis. In production, the fiber is generally fusion spliced.

Direct modulation of DBR and GCSR lasers is possible at OC-48 rates. It is also possible to integrate electroabsorption modulators with these structures or to create hybrid integrated semiconductor external modulators. In principle, VCSEL lasers can also be directly modulated.

Wavelength stability and switching

The key to effective use of tunable lasers in networks is that the wavelength must remain stable over time. For example, a laser module could be specified to be within 3 GHz of the desired wavelength over all operating environmental conditions. To maintain this accuracy over the module`s lifetime, it is necessary, as with a DFB laser, to use an external wavelength reference.

The principle of wavelength locking is simple. Part of the optical signal is tapped off and put through a wavelength-selective passive component (typically an etalon). This signal is compared to a reference signal and to a feedback signal generated if the wavelength is incorrect. The feedback can be performed directly with the electrical feedback output of commercial wavelength lockers or through modules that incorporate separate lockers (see Fig 3).

It should be remembered that DBR and GCSR type lasers can tune to any wavelength in their tuning range, whereas lockers are generally periodic. This function can be a limitation, as commercial lockers now generally operate on just a 100-GHz grid (either on the ITU grid or at a 50-GHz offset). The locker is therefore an additional limitation on the tuning ability, making an argument for keeping the locker external to the laser package initially.

Finally it should be noted that present commercial tunable laser modules are designed to switch wavelengths in around 1 ms. Fast packet switching of wavelengths in the nanosecond range is possible, but special electronics are required to rapidly inject current into the tuning sections.2

Tunable lasers offer many advantages in both long-haul and metro networks, and they are rapidly entering commercial applications. Modules incorporating all required control electronics allow these lasers to be rapidly designed into systems and to be readily interchangeable.

REFERENCES

1. "Experimental Demonstration of a Media Access Protocol for HORNET: A WDM Multiple Access Metropolitan Area Ring Network," I. M. White et al., OFC 2000 Technical Digest Paper WD3.

2. "Fast and fine wavelength tuning of a GCSR laser using a digitally controlled driver," Y. Fukashiro et al., OFC 2000 Technical Digest Paper WM43.

FIGURE 3. In a typical arrangement using a tunable laser module the sample signal is tapped after the modulator. It is possible to bring the locking optics inside the laser package, which is the long-term solution.